This system allow recognition the number plates on cars.

The system is compsed of the following elements:

Two input device: a CCD camera to capture image and a photocell to trigger the process;

Two interface cards for PC: an I/O port and a video capture card;
A computer running Windows 95/98/NT/2000/XP, that hosts the application core and application software;
The application core the software acquisition and recognition
The application software: the custom software, the end user applications
Some accessories: lamps to enhance the image acquisition, support to hold the components, PC housing units

The operation is performed in several steps of execution:

A input trigger (for example a pulse from the photocell) start the acquisition process;
The CCD camera, using a infrared filter, acquires the image at 256 gray-scale level, that will be processed by the system.
The system find the number plate position into picture, discarding the boundary elements such as car, road, people, billboards, etc...
The number plate image, so extracted, is transformed into a monochrome format, using a process of binarization with dynamic tresholding;
Each character of the plate is recognized using a neural network classifier, specifically trained on this type of shape;
The recognized string, undergoes a process of verify using a vector of pattern allowed;
The result string, extracted whit a confidence interval, and the picture recognized is send to the application software;
This last can verify in the presence of a list of the number of plate and decides whether to perform an action such as open a bar or a gate, operate a traffic light, print a ticket ...
The performace are excellent, although the procedure is complex: the system takes less than 5 seconds to reconize the plate with an accuracy that exceeds 95% of exact matches



Only for V.I.P